GNN Masterclass with Graphcore Research

Now Available On Demand

Graph Neural Networks (GNNs) are taking the AI community by storm. Gaining ground for applications such as drug discovery, GNNs are well suited to representing countless real-world phenomena featuring irregular structures, from proteins to social networks. 

As presented at the NeurIPS OGB Workshop, following a double-win at the Open Graph Benchmark Large Scale Challenge (OGB-LSC 2022), our researchers are inviting you to learn more about these two award-winning graph models: Distributed KGE for predicting missing facts in a knowledge graph and GPS++ for molecular property prediction.

You’ll also learn more about Graph Neural Networks and how to accelerate them for real-world applications. 

Plus, we’ll feature a use case on molecular modelling from Pacific Northwest National Laboratory and provide real-world performance insights with state-of-the-art IPU technology.

Takeaways for your own GNN projects:

• What Graph Neural Networks are and how you can apply them to real-world applications
• How to accelerate GNNs with easy-to-implement technologies and performance tips
• Understanding and using OGB-winning models: GPS++ (from Graphcore, MILA & Valence Discovery) and Distributed KGE
• Real-world use case: how PNNL accelerated molecular GNN training with IPUs from about 3 days to just over 1 hour
• Try GNNs for free: how to get a free cloud trial to test GNNs on IPUs